10.

5 SINGLE PERIOD INVENTORY MODEL

WITH PROBABILISTIC DEMAND
 The inventory models discussed this far were based on the assumption that
the demand rate is constant and deterministic throughout the year. We
developed minimum cost order quantity and reorder point policies based on this
assumption. In situations for which the demand rate is not deterministic, other
models treat demand as probabilistic and best described by a probability
distribution. In this section we consider a single-period inventory model with
probabilistic demand.

The single-period inventory model refers to inventory situations for which

one order is placed for the product; at the end of the period, the product has either
sold out, or a surplus of unsold items will be sold for a salvage value. The
single-period inventory model is applicable in situations involving seasonal or
perishable items that cannot be carried in inventory and sold in future periods.
Seasonal clothing (such as bathing suits and winter coats) are typically handled in
a single-period manner.
 In these situations, a buyer places one preseason order for each item and
then experiences a stock-out or holds a clearance sale on the surplus stock at the
end of the season. No items are carried in inventory and sold the following year.
Newspapers are another example of a product that is ordered one time and is
either sold or not sold during the single period. Although newspapers are ordered
daily, they cannot be carried in inventory and sold in later periods. Thus,
newspaper orders may be treated as a sequence of single-period models; that is,
each day or period is separate, and a single-period inventory decision must be
made each period (day). Because we order only once for the period, the only
inventory decision we must make is how much of the product to order at the start
of the period.

Obviously, if the demand were known for a single-period inventory situation,

the solution would be easy; we would simply order the amount we knew would
demanded.
 However, in most single-period models, the exact demand is not known. In
fact, forecasts may show that demand can have a wide variety of values. If we are
going to analyze this type of inventory problem in a quantitative manner, we need
information about the probabilities associated with the various demand values.
Thus, the single-period model presented in this section is based on probabilistic
demand.

Neiman Marcus

Let us consider a single-period inventory model that could be used to make a

how-much-to-order decision for Neiman Marcus, a high-end fashion store. The
buyer for Neiman Marcus decided to order Manolo Blahnik heels shown at a buyers’
meeting in New York City. The shoe will part of the company’s spring-summer
promotion and will be sold through nine retail stores in the Chicago area. Because
the shoe is designed for spring and summer months, it cannot be
expected to sell in the fall. Neiman Marcus plans to hold a special August
clearance sale in an attempt to sell all shoes not sold by July 31. The shoes cost
$700 a pair and retail for $900 a pair. At the sale price of $600 a pair, all surplus
shoes can be expected to sell during the August sale. If you were the buyer for
Neiman Marcus, how many pairs of the shoes would you order?

To answer the question of how much to order, we need information on the

demand for the shoe. Specifically, we we would need to construct a probability
distribution for the possible values of demand. Let us suppose that the uniform
probability distribution shown in Figure 10.8 can be used to describe the demand
for the Manolo Blahnik heels.
Figure 10.8 UNIFORM PROBABILITY DISTRIBUTION OF DEMAND FOR
NEIMAN MARCUS PROBLEM
Expected
Demand = 500

350 500 650

DEMAND
 In particular, note that the range of demand is from 350 to 650 pairs of
shoes, with an average, or expected, demand of 500 pairs of shoes.

Incremental analysis is a method that can be used to determine the optimal

order quantity for a single-period inventory model. Incremental analysis addresses
the how-much-to-order question by comparing the cost or loss of ordering one
additional unit with the cost or loss of not ordering one additional unit. The costs
involved are defined as follows:

Co = cost per unit of overestimating demand. This cost represents the loss of
ordering one additional unit and finding that it cannot be sold.

Cu = cost per unit of underestimating demand. This cost represents the

opportunity loss of not ordering one additional unit and finding that it could have
been sold.
 In the Neiman Marcus problem, the company will incur the cost of
overestimating demand whenever it orders too many pairs and has to sell the
extra shoes during the August clearance sale. Thus, the cost per unit of
overestimating demand is equal to the purchase cost per unit minus the August
sales price per unit; that is Co = $700 - $100. Therefore, Neiman Marcus will lose
$100 for each pair of shoes that it orders over the quantity demanded. The cost of
underestimating demand is the lost profit (often referred to as an opportunity
cost) because a pair of shoes that could have been sold was not available in
inventory. Thus, the per-unit cost of underestimating demand if the difference
between the regular selling price per unit and the purchase cost per unit; that is
Cu = $900 - $700 = $200.
Because the exact level of demand for the Manolo Blahnik heels is unknown, we
have to consider the probability of demand and thus the probability of obtaining
the associated costs or losses. For example, let us assume that Neiman Marcus
management wishes to consider an order quantity equal to the average or
expected demand for 500 pairs of shoes. In incremental analysis, we consider the
possible losses associated with an order quantity of 501 (ordering one additional
unit) and an order quantity of 500 ( not ordering one additional unit). To order
quantity alternatives and the possible losses are summarized here.

Note:

The key to an incremental analysis is to focus on the costs that are different when
comparing an order quantity Q + 1 to an order quantity Q.
Order Quantity Loss Occurs Possible Loss Probability Loss
Alternatives If Occurs

Demand
Q = 501 overestimated; the Co = $100 P(demand < 500)
additional unit
cannot be sold

Demand
Q = 500 underestimated; the Cu = $200 P(demand > 500)
additional unit
could have been
sold
 Using the demand probability distribution in Figure 10.8, we see the
P(demand < 500) = 0.50 and that P(demand > 500) = 0.50. By multiplying the
possible losses, Co = $100 and Cu = $200, by the probability of obtaining the
loss, we can compute the expected value of the loss, or simply the expected loss
(EL), associated with the order quantity alternatives. Thus,

EL(Q = 501) = Co P(demand < 500) = $100(0.50) = $50

EL(Q = 500) = Cu P(demand > 500) = $200(0.50) = $100

Based on these expected losses, do you prefer an order quantity of 501 or

500 pairs of shoes? Because the expected loss is greater for Q = 500, and
because we want to avoid this higher cost or loss, we should make Q = 501 the
prefered decision. We could now consider incrementing the order quantity one
additional unit to Q = 502 and repeating the expected loss calculation.
 Although we could continue this unit-by-unit analysis, it would be time
consuming and cumbersome. We would have to evaluate Q = 502, Q = 503, Q =
504 and so on until we found the value of Q where the expected loss of ordering
one incremental units is equal to the expected loss of not ordering one
incremental unit; that is, the optimal order quantity Q* occurs when the
incremental analysis shows that
EL(Q* + 1) = EL(Q*) (10.29)

When this relationship holds, increasing the order quantity by one additional unit
has no economic advantage. Using the logic with which we computed the
expected losses for the order quantities of 501 and 500, the general expressions
for EL (Q* + 1) and EL (Q*) can be written as

EL(Q*+1) = Co P(demand ≤ Q*) (10.30)

EL(Q*) = Cu P(demand > Q*) (10.31)
Because demand ≤ Q* and demand > Q* are complementary events, we know
from basic probability that

P(demand ≤ Q*) + P(demand > Q*) = 1 (10.32)

and we can write

P(demand > Q*) = 1 - P(demand ≤ Q*) (10.33)

Using this expression, equation (10.31) can be rewritten as

EL(Q*) = Cu [1 - P(demand ≤ Q*] (10.34)

Equations (10.30) and (10.34) can be used to show that EL(Q* + 1 ) = EL (Q*)
whenever

Co P(demand ≤ Q*) = Cu [1 - P(demand ≤ Q*] (10.35)

Solving for P(demand ≤ Q*), we have

P(demand ≤ Q*) = Cu (10.36)

Cu+Co
This expression provides the general condition for the optional order quantity Q* in
the single-period inventory model.
In the Neiman Marcus problem, Co = $100 and Cu = $200. Thus, equations
(10.36) shows that the optimal order size for the Manolo Blahmik heels must satisfy
the following condition:

P(demand ≤ Q*) = Cu = 200 = 200 = 2

Cu+Co 200 + 100 300 3
 We can find the optimal order quantity Q* by referring to the probability
distribution shown in Figure 10.8 and finding the value of Q* that will provide
P(demand ≤ Q*) = ⅔. To find this solution, we note that in uniform distribution the
probability is evenly distributed over the entire range of 350-650 pairs of shoes.
Thus, we can satisfy the expression for Q* by moving two-thirds of the way from
350 to 650. Because this range is 650 - 350 = 300, we move 200 units from
350 toward 650.

Range of Possible Sales

I I I I I I I I I
300 350 400 450 500 550 600 650 700

⅔ of the Range of Possible Sales

Doing so provides the optimal order quantity of 550 pairs of shoes.
In summary, the key to establishing an optimal order quantity for
single-period inventory models is to identify the probability distribution that
describes the demand for the item and to calculate the per-unit costs of
overestimation and underestimation. Then, using the information for the per-unit
costs of overestimation and underestimation, equation (10.36) can be used to find
the location of Q* in the probability distribution.

Nationwide Car Rental

As another example of a single-period inventory model with probabilistic demand,
consider the situation faced by National Car Rental. Nationwide must decide how
many automobiles to have available at each car rental location at specific points in
time throughout the year. Using the Myrtle Beach, South Carolina, location as an
example, management would like to know the number of full-sized automobiles to
have available for the Labor Day weekend.
Based on previous experience, customer demand for full-sized automobiles to
have available for the Labor Day weekend has a normal distribution with a mean
of 150 automobiles and a standard deviation of 14 automobiles.
The National Car Rental situation can benefit from use of a single-period
inventory model. The company must establish the number of full-sized
automobiles to have available prior to the weekend. Customer demand over the
weekend will then result in either a stock-out or surplus. Let us denote the
number of full-sized automobiles available by Q. If Q is greater than customer
demand, Nationwide will have a surplus of cars. The cost of a surplus is the cost
of overestimating demand. This cost is set at $80 per car, which reflects, in part,
the opportunity cost of not having the car available for rent elsewhere. If Q is less
than the customer demand, Nationwide will rent all available cars and experience
a stock-out or shortage. A shortage results in an underestimation cost of $200 per
car. This figure reflects the cost due to lost profit and the lost goodwill of not
having a car available for customer.
Given this information, how many full-sized automobiles should Nationwide make
available for Labor Day weekend?
Using the cost underestimation, Cu=$200, and the cost of overestimation
Co=$80, equation (10.36) indicates that the optimal quantity must satisfy the
following condition:

P (demand ≤ Q*) = Cu = 200 = 0.7143

Cu + Co 200 + 80
Figure 10.9 PROBABILITY DISTRIBUTION OF DEMAND FOR THE NATIONWIDE CAR
RENTAL PROBLEM SHOWING THE LOCATION OF Q*

P (demand ≤ Q*) = 0.7143 σ = 14

150
Q* = 158
 We can use the normal probability distribution for demand as shown in Figure
10.9 to find the order quantity that satisfies the condition that P (demand ≤ Q*) =
0.7143. From appendix B, we see that 0.7143 of the area in the left tail of the
normal probability distribution occurs at z = 0.57 standard deviation above the
mean. With a mean demand of μ = 150 automobiles and a standard deviation of
σ = 14 automobiles, we have

Q* = μ + 0.57σ
= 150 + 0.57 (14) = 158

Thus, Nationwide Car Rental should plan to have 158 full-sized automobiles
available in Myrtle Beach for the Labor Day weekend. Note that in this case, the
cost of overestimation is less than the cost of underestimation. Thus, Nationwide
is willing to risk a higher probability of overestimating demand and hence a higher
probability of a surplus. In fact, Nationwide’s optimal order quantity has 0.7143
probability of a surplus and 1 - 0.7143 = 0.2857 probability of a stock-out. As a
result, the probability is 0.2857 that all 158 full-sized automobiles will be rented
during the Labor Day weekend